[Copied to [EMAIL PROTECTED]] Circa 2002-Jun-25 11:31:04 +0900 dixit [EMAIL PROTECTED]:
: >> +# use libcrypt if there is : >> +AC_CHECK_LIB(crypt, crypt, LIBS="$LIBS -lcrypt") : >> + : >AC_CHECK_LIB(crypt, crypt) will automatically add -lcrypt to $LIBS. It : >will also define HAVE_LIBCRYPT (is this what you're trying to avoid)? : >Anyway, I'd prefer: : > AC_CHECK_FUNCS(crypt, , AC_CHECK_LIB(crypt, crypt)) : >This way we check if crypt is resolvable using the existing $LIBS and, : >if not, use $LIBS+-lcrypt. : : either way is fine for me, as long as crypt() supplied by the : native system is preferred than openssl crypt(). thanks. : : itojun Isn't this really a problem for OpenSSL? I know that several vendors (notably Linux ones...) already patch OpenSSL to remove crypt() from OpenSSL's libcrypto, so that crypt() is only available via the system libcrypt. Even the stock OpenSSL-0.9.6d sources omit crypt() under FreeBSD, NeXT, and Darwin. I really think that OpenSSL should not contain crypt() at all. For situations where the system crypt() is so broken as to prefer OpenSSL's implementation, the symbol should be openssl_crypt(), or something similarly named, and it's up to the calling application to #define it as crypt() or not. -- jim knoble | [EMAIL PROTECTED] | http://www.pobox.com/~jmknoble/ (GnuPG fingerprint: 31C4:8AAC:F24E:A70C:4000::BBF4:289F:EAA8:1381:1491)
msg26471/pgp00000.pgp
Description: PGP signature
